Essays on optimal inflation targeting forecast based rules and inflation modelling under uncertainty

This thesis focuses on exploring the most efficient forecast based rules for Inflation Targeting (IT) regime; modelling and forecasting inflation. Therefore, it is divided into four empirical chapters after the introduction as the first chapter follows the second chapter explores the most efficient alternative forecast based rules in the context of an estimated Global Projection Model for the Egyptian Central Bank’s IT period. In addition to the traditional Inflation Forecast Based (IFB) rule, the chapter augments this rule with other variables like the expected exchange rate or output gap or both. Also, it proposes a structural Backward-Forward rule (BF) which implies more dynamics for the monetary policy rule and it encompasses other common rules. The third chapter discusses modelling and forecasting inflation from Phillips Curve (PC) under misspecification. It considers various econometric specifications, estimation methods, and different measures of business cycles. Then, we propose a Time Varying Coefficient Phillips Curve (TVCPC) which is more sophisticated and informative and, also, it acts as a tool to make the gap between the actual specification and the estimated one as small as possible. The fourth chapter considers: modelling the density of quarterly inflation by using a time-varying higher order moment’s model developed by Leon, Rubio, and Serna (2005); and isolating the time-varying conditional correlations between inflation and both the growth in domestic credit and the real exchange rate by using two multivariate GARCH models. The fifth chapter focuses on improving inflation forecasts through combining some linear and non-linear models by using both traditional and other proposed sophisticated time varying combination approaches. We find that the BF rule is the superior welfare policy under all policy scenarios. With regard to the IFB rule, we conclude that: the augmented versions with the expected exchange rate are preferable to the IFB rule; the Time Varying Coefficient Phillips Curve with HP output gap (TVCPC_HP) produces the best forecasting accuracy; and models with time-invariant volatility, skewness and kurtosis are inferior to the models with time-varying higher order moments. Moreover, in comparison to static models, dynamic multivariate models can provide rich information related to inflation dynamics and forecasts. Further, the proposed time varying combination approaches dominate all individual models and all other static combination schemes.